Open JagatguruMishra opened 1 month ago
There are some more details of the solution / workaround regarding this huge_pages
issue, here:
The huge_pages = off
configure might need to be apply to this file with code change similar like below:
sed -i -r 's/#huge_pages.*?/huge_pages = off/g' /usr/pgsql/15/share/postgresql/postgresql.conf.sample
In that case, the huge_pages
will be turned off to all users, even those who want huge_pages
to be try
or on
. Having huge_pages = off
may have performance issue. As it states here:
The use of huge pages results in smaller page tables and less CPU time spent on memory management, increasing performance.
Thanks @zyyw. I am using the image goharbor/harbor-db:v2.8.0, so correct file is /usr/pgsql/13/share/postgresql/postgresql.conf.sample. I don't prefer to build an image due to maintenamce concers, also I can't disable hugepages on the host system or use the available huge pages. I am looking for a possibility where I can disable hugepages by modifying the statefulset, may be by adding any env value or running some commands in init container or any other way to configure it.
Hello @JagatguruMishra, I'm currently facing the same issue and looking into not having to modify the image nor disabling hugepages on the host system, were you able to find a suitable solution? Thanks.
Hi @gera-aldama I was able to apply below workaround.
I am installing harbor with helm chart version 1.12.0 (https://github.com/goharbor/harbor-helm/archive/refs/tags/v1.12.0.tar.gz). Pod 'harbor-database-0' is going in CrashLoopBackOff status and some other pods dependent on this are going to the same state. Below are the logs.
The issue is with the huge pages. If I provide the hugepages resource to the pod like 'hugepages-1Gi: 1' or disable the huge pages on the node, it works.
Is there any way to provide the configuration in harbor-db pod/statefulset to disable hugepages (setting huge_pages = off ) in .conf file.